Performance Comparison of Deep Learning Models in Detecting Tuberculosis and Pneumonia from Chest X-Rays

Both tuberculosis (TB) and pneumonia are serious infectious illnesses that primarily affect the respiratory system and have a considerable influence on world health. Acute bacterial, viral, or fungal infection that causes fever, coughing, and dyspnea, pneumonia primarily affects young children, the elderly, and people with weakened immune systems. Mycobacterium tuberculosis, the cause of tuberculosis (TB), is a long-term illness that spreads by airborne particles and is characterized by a persistent cough, fever, night sweats, and weight loss. Critical public health issues are raised by both illnesses, particularly in low- and middle-income nations. The genesis, transmission, clinical manifestation, and public health consequences of pneumonia and tuberculosis are examined in this work. In order to effectively lessen their global impact, emphasis is placed on the need for comprehensive strategies in prevention, diagnosis, and treatment that are in accordance with the sustainable development goals of good health and well-being, reduced inequality, no poverty, quality education, clean water and sanitation, and partnerships for the goals. This paper applies four distinct models, namely VGG16, VGG19, ResNet50, and EfficientNetB0. It is observed that the EfficientNetB0 model outperforms the other models. The EfficientNetB0 model had a peak accuracy of 0.9840 during the training phase and 0.9635 during the validation phase. The training loss is 0.00451, whereas the validation loss is 0.1339.